This seems to come down to object model and tooling differences between sandboxed JavaScript and sandboxed Unix. You can make Unix commands do whatever you like, including special commands to send things back to the coding agent. But the output of a Unix command is arbitrary text (by default) and the AI will have to pipe things into some other command when there’s a lot of output. In the coding harness I use, I see OpenAI’s LLMs writing a lot of pipelines, often processing the output with python. Also the whole thing needs to run in a VM.

A JavaScript sandbox works somewhat differently since the output is an object. Tools are functions that naturally return objects. Composing functions and async calls work differently. If a function returns a very large object, the coding harness could be smarter about presenting the result to the LLM. Maybe JacaScript works better than bash for calling mcp APIs?

But you can have both! The LLM can get direct access to a JavaScript sandbox, which in turn provides an API to access a Linux sandbox. This seems to be what pi is doing with code mode?